strange problem ahout IIP3

[tangqin55]

ip3 slope is not 3

When I simulated IIP3 of a Double-balanced CMOS Mixer, I find the slope is not 3 but about 2.35, so selecting different input amplitude to extrapolate, the IIP3 result is different. Why dose this instance occur? When meeting this problem, how can I simulate IIP3?

 

[normaldist]

It could be the amplitude of the signal you are using for your simulation. When the amplitude is high it is possible for higher order IM to intefere with your measurment. IPs need to be measured at different power levels to insure that the reading is consistant. The calculated IP3 should be the same for at least two power levels. Too often people measure this at too high a power level and get an IP3/IP5/IP7 combo.

C

 

[tangqin55]

I have took it into consideration, so I set amp_RF=-60dBm and amp_LO=0dBm,but the result was still strange which can be seen in the picture. I think the input amplitudes is small enough, isn't it?

 

[RFCMOS]

Hi

I don't perhaps contribution of the 5th order non-linearity....

 

[tangqin55]

fRF=432MHz,fLO=396MHz,fIF=36MHz.so the frequence of IM3 are 35MHz and 37MHz, and the frequence of IM5 are 34MHz and 38MHz. How should I do to identify where the reason is IM5?

 

[Mazz]

From your sim resutls it seems that the proper input power to evaluate the IP3 is below -20 dBm.
Take care of accuray settings: decrease by two order of mag the 3 main analog accuracy settings from default values. You will need a larger time for sim, but with more reliable results.
I hope it can help.
Mazz

 

[tangqin55]

"decrease by two order of mag the 3 main analog accuracy settings from default values"
I"m sorry I can t understand it clearly,could you tell me in detail?
If I choose the extrapolate point below -20dBm, the IIP3 will de very small. I think the actual curve of IM3 should below the ideal curve when the input amplitude is very small(amp_RF=-60dBm is small enough, isn't it?), but unfortunately, the sim result is not.Why?

 

[Mazz]

I will try to explain better:
From the plot you are showing us, it is clear that you are using Spectre RF, PSS and Pac. My opinion is that the strange result you are getting is due to inaccuracy of simulation settings.
First issue to check are the analog accuracy settings, such as reltol, abstol and so on. To proper simulate IP3 you should decrease these settings (for example reltol default value=1e-3 -> use 1e-5).
If, after that, the simulation is still giving you wrong results, so use two large signals at Mixer Linear port instead of one large and one small signal, avoiding the Pac analysis. Let me advise you that this can be a very heavy simulation.
I hope I've been more clear, now.
Mazz

 

[strabush]

All the simulators that depend on BSIM mosfet models have an inherent problem in the mosfet model. there is an discontinuity in the derivative of the dId/dVds when Vds crosses 0 volts. This is not happening in regular amplifier circuits but only in mosfets used as switches, and especially in passive mixers. There is no clear cut solution for the problem. The BSIM comitee is working on it but no solution yet.

 

[tangqin55]

Thanks everybody! I have made it

 

[Ben_edaboard]

How did you make it? Could you share it?